Distributed Plant Hydraulic and Hydrological Modelingto Understand the Susceptibility of Riparian WoodlandTrees to Drought-Induced MortalityXiaonan Tai1 , D. Scott Mackay1 , John S. Sperry2, Paul Brooks3 , William R. L. Anderegg2,Lawrence B. Flanagan4 , Stewart B. Rood4, and Christopher Hopkinson5

1Department of Geography, State University of New York at Buffalo, Buffalo, NY, USA, 2Department of Biology, University ofUtah, Salt Lake City, UT, USA, 3Department of Geology and Geophysics, University of Utah, Salt Lake City, UT, USA,4Department of Biological Sciences, University of Lethbridge, Lethbridge, Alberta, Canada, 5Department of Geography,University of Lethbridge, Lethbridge, Alberta, Canada

Abstract The mechanistic understanding of drought-induced forest mortality hinges on improvedmodels that incorporate the interactions between plant physiological responses and the spatiotemporaldynamics of water availability. We present a new framework integrating a three-dimensional groundwatermodel, Parallel Flow, with a physiologically sophisticated plant model, Terrestrial Regional EcosystemExchange Simulator. The integrated model, Parallel Flow-Terrestrial Regional Ecosystem Exchange Simulator,was demonstrated to quantify the susceptibility of riparian cottonwoods (Populus angustifolia, Populusdeltoides, and native hybrids) in southwestern Canada to sustained atmospheric drought and variability instream flow. The model reasonably captured the dynamics of soil moisture and evapotranspiration in bothwet and dry years, including the resilience of cottonwoods despite their high vulnerability to xylemcavitation. Unrealistic predictions of mortality could be generated when ignoring lateral groundwater flow.Our results also illustrated a mechanistic linkage between streamflow and cottonwood health. In the absenceof precipitation, normal streamflow could sustain 94% of cottonwoods, and higher streamflows would berequired to sustain all of the floodplain cottonwoods. Further, the risk of mortality was mediated by planthydraulic properties. These results underpin the importance of integrating groundwater processes and planthydraulics in order to analyze the forest response to sustained severe drought, which could increase in thefuture due to climate change combined with increasing river water withdrawals.

1. Introduction

Rapidly changing environmental conditions have recently exposed vegetation to extreme drought andcaused increased mortality in many ecosystems across the globe (Allen et al., 2015). The impacts of droughton ecosystems can be diverse (Bond et al., 2008), and it is challenging for ecosystem models to predict plantresponses to drought (Powell et al., 2013). This can be attributed in part to oversimplified representations ofboth plant responses to soil drying (Fisher et al., 2017; McDowell et al., 2013) and heterogeneous plant watersupply across the landscape (J. S. Clark et al., 2016; Fang et al., 2017).

Current models fail to predict plant response to drought (Hanson et al., 2004; Powell et al., 2013), as they lar-gely rely on simple stress functions with parameters empirically fitted to historic nondrought conditions(Sperry et al., 2016; Xu et al., 2013). Relevant physiological mechanisms are needed for models to reliably pre-dict plant responses to environmental change (McDowell et al., 2013). Plant hydraulics has emerged as a keymechanism for understanding plant water relations (Sack et al., 2016; Venturas et al., 2017), especially underextreme drought (Sperry & Love, 2015). Plant hydraulics provides mechanistically grounded descriptions ofwater transport along the soil-plant-atmosphere continuum based on measurable plant and soil hydraulicproperties (Jackson et al., 2000; Manzoni et al., 2013; Sperry & Love, 2015). Incorporating plant hydraulicshas been shown to improve predictions of tree response to drought stress across different species and acrossthe landscape (W. R. Anderegg et al., 2015; Mackay et al., 2015; McDowell et al., 2013; Tai et al., 2017). The suc-cess of these studies argues for incorporating plant hydraulics into regional models to gain insights on forestdrought responses at the landscape scale (Jackson et al., 2000; Sperry et al., 2016).

Advances in understanding plant response to atmospheric drought highlight the need for improvements indescribing belowground processes that mediate water availability (Billings, 2015; Fang et al., 2017; Phillips

TAI ET AL. 4901

Water Resources Research

RESEARCH ARTICLE10.1029/2018WR022801

Key Points:• Plant hydraulics and hydrology are

integrated• Role of alternate water sources in

sustaining cottonwoods is assessed• Susceptibility to different

streamflows is predicted

Supporting Information:• Supporting Information S1• Data Set S1

Correspondence to:X. Tai,xiaonant@buffalo.edu

Citation:Tai, X., Mackay, D. S., Sperry, J. S.,Brooks, P., Anderegg, W. R. L., Flanagan,L. B., Rood, S. B., & Hopkinson, C. (2018).Distributed plant hydraulic andhydrological modeling to understandthe susceptibility of riparian woodlandtrees to drought-induced mortality.Water Resources Research, 54, 4901–4915.https://doi.org/10.1029/2018WR022801

Received 16 FEB 2018Accepted 23 JUN 2018Accepted article online 2 JUL 2018Published online 21 JUL 2018

©2018. American Geophysical Union.All Rights Reserved.

et al., 2016). In addition to precipitation infiltration from the top, soilscan also be wetted from below or laterally through groundwater pro-cesses at local, landscape, and regional scales (Figure 1). At local scales,the water stored in deep soil can sustain plants through multimonthdry seasons after being recharged in the wet season (Miguez-Macho& Fan, 2012; Newman et al., 2006). At landscape scales, groundwaterredistributes along topographic gradients to form areas that are consis-tently wetter than others (Beven & Kirkby, 1979; Dingman, 1994; Fan,2015; Western et al., 1999). Those wetter areas recently have beenshown to support greater productivity and tree growth across a broadrange of climates (Swetnam et al., 2017; Thompson et al., 2011). Atregional scales, snowmelt from higher elevations supports streamflowsand provides remote water sources for the development of riparianvegetation in the semiarid regions (Barnett et al., 2005; Dawson &Ehleringer, 1991).

Groundwater operates across a range of spatial and temporal scales(Schaller & Fan, 2009; Toth, 1963) and can provide sources of plantwater that are decoupled and/or asynchronous with the timing of pre-cipitation (Fan, 2015). The consequence of ignoring groundwaterhydrology is therefore a failure to account for the different droughtintensities experienced by trees within the same or similar climatic con-ditions, leading to underestimation of the spatial variability of droughtimpacts and vegetation survival at landscape scales (Fisher et al., 2017).Given its potentially large storage and slow response to climate change(Taylor et al., 2013), groundwater can be critical for mediating theimpact of extreme drought on plants and for promoting ecological

refugia and seed banks in topographically low areas or along river corridors (Fan, 2015; Krause et al., 2017).

Riparian ecosystems are simultaneously influenced by groundwater processes from local to regional scales(Bergstrom et al., 2016; Cardenas, 2015; Figure 1). Cottonwoods (Populus spp.) are a predominant speciesin riparian ecosystems throughout arid and semiarid regions of North America (Rood et al., 2013). Theyare expected to be particularly susceptible to the anticipated future droughts given their high vulnerabil-ity to xylem cavitation (Tyree et al., 1994) and high sensitivity to declining streamflows (Rood et al., 2013;Scott et al., 1999). Efforts to manage and maintain existing cottonwood forests require quantifying theinteraction between their survival and groundwater dynamics (Amlin & Rood, 2003; Flanagan et al.,2017; Scott et al., 1999). Here we present a framework that combines the most recent knowledge of planthydraulics and variably saturated groundwater modeling to quantify the risk of drought-induced treemortality across space and time. This integrated model includes much more complete physics-basedrepresentations than current approaches and is expected to be more robust when applied to novel envir-onmental conditions (M. P. Clark et al., 2015). By quantifying the mortality risk of a cottonwood ecosystemunder different scenarios, we evaluated the following two hypotheses: (1) increasing plant water sourcesbuffers ecosystem vulnerability to drought and (2) ecosystem vulnerability can be further mediated byplant hydraulic traits.

2. Materials and Methods2.1. Integrated Modeling

An integrated model was developed by coupling a plant physiology model, Terrestrial Regional EcosystemExchange Simulator (TREES; Mackay et al., 2015), to a variably saturated groundwater model, PARallelFLOW (ParFlow; Kollet & Maxwell, 2008). TREES has been used to successfully predict tree hydraulic stressand mortality (Johnson et al., 2018; Mackay et al., 2015; McDowell et al., 2013; Tai et al., 2017). It solves planttranspiration and photosynthesis based on meteorological forcing of temperature, precipitation, photo-synthetically active radiation, vapor pressure deficit, wind speed, atmospheric CO2 concentration, and atmo-spheric pressure. Here TREES was redesigned around a parsimonious model based on plant hydraulics theory

Figure 1. Conceptual diagram showing the contribution of groundwater pro-cesses to mediate plant water supply and risk of drought-induced mortality.The x axis represents the cumulative plant water sources, with small-scale flownested inside large-scale flow. The y axis represents the ecosystem level mor-tality risk. Here we simplified groundwater processes into three categories basedon the spatial scales of water sources. At local scales, we refer to the vertical flowin the root zone or deeper soil. At landscape scales, we refer to the lateral flowdriven by topographic gradients in the absent of long-distance regional flow. Atregional scales, we refer to long-distance streamflow driven by large-scale cli-matologic and geologic factors, as well as stream management along its trans-portation. Furthermore, plant hydraulic strategies (shown by the differentlycolored curves) mediate the risk of mortality to a given water supply withdrought vulnerable plants requiring a greater water supply to be buffered fromdrought stress.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4902

to predict transpiration (Sperry & Love, 2015; Sperry et al., 2016). This facilitated its integration with adistributed hydrological model, ParFlow, which solves both the surface and subsurface groundwatersystems simultaneously. ParFlow is well documented and has also been coupled with the CommunityLand Model (Kollet & Maxwell, 2008).

In the coupled ParFlow-TREES model, the coupling takes place through soil matric potential and thesource/sink terms of evapotranspiration (ET) and infiltration from precipitation. The soil water scheme inthe original TREES was replaced with ParFlow, which solves the profile of the soil water potential at each gridcell in the computation domain. TREES derives water uptake from each root layer and feeds this flux back toParFlow to update the local soil water potential at every time step. The integratedmodel, ParFlow-TREES, cap-tures the water movement within soils, and along the soil-plant-atmospheric continuum, following waterpotential gradients both in the lateral directions within the subsurface and vertically within the soil-plant-atmosphere continuum (Figure 2). ParFlow-TREES can be driven by parameters that are spatially uniformor heterogeneous and takes advantage of parallel computation.

Complete details on the equations and parameters in TREES and ParFlow can be found in previous literature(Kollet & Maxwell, 2006, 2008; Mackay et al., 2015; Maxwell & Kollet, 2008; Sperry & Love, 2015; Sperry et al.,2016), and brief summary of the relevant equations is provided here. In the subsurface, ParFlow solves themixed form of Richards’ equation for variably saturated flow (Richards, 1931) in the three spatial dimensionsfollowing the notion in (Maxwell et al., 2015)

SsSw hð Þ ∂h∂t

þ φSw hð Þ ∂Sw hð Þ∂t

¼ ∇·qþ qr=λ; (1a)

where the flux term q (m/hr) is based on Darcy’s law:

q ¼ �K s xð Þkr hð Þ ∇ hþ zð Þ cosθx þ sinθx½ �; (1b)

where h is the pressure head (m), z is the elevation (m), x is the index of grid cells in the domain, Ks(x) is thesaturated hydraulic conductivity tensor (m/hr), kr(h) is the relative permeability, Ss is the specific storage(m�1), φ is the porosity, Sw(h) is the relative saturation, qr is a source/sink term (e.g., precipitation, ET; m/hr)with λ being the thickness of the surface layer (m), and θ is the local slope angle in the lateral directions.The van Genuchten relationships (Van Genuchten, 1980) are used to describe the relative saturation and per-meability functions Sw(h) and kr(h).

Overland flow is represented by the two-dimensional kinematic wave equation (Kollet & Maxwell, 2006):

Figure 2. Conceptual diagram of the integrated model coupling a plant physiology model, Terrestrial Regional EcosystemExchange Simulator to a variably saturated groundwater model, Parallel Flow. The integrated model, Parallel Flow-Terrestrial Regional Ecosystem Exchange Simulator, captures the three-dimensional water movement in the subsurface andvertically through the soil-plant-atmosphere continuum. It allows the groundwater to laterally redistribute from topo-graphic high to low or from stream channel into the floodplain, driven by the hydraulic gradients.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4903

�K s xð Þkr hð Þ·∇ hþ zð Þ ¼ ∂ h; 0k k∂t

� ∇· max h; 0k kvsw þ qr ; (2)

where vsw is the two-dimensional, depth-averaged surface-water velocity (m/hr) given by Manning’s equa-tion (Chow et al., 1988); max‖h, 0‖ indicates the greater of h and 0. This results in the kinematic wave equationbeing only active when the pressure at the top cell of the domain is greater than 0 (h > 0), and this pondedwater is routed following pressure gradients (Kollet & Maxwell, 2006).

The nonlinear, coupled equations of surface and subsurface flow are solved using a parallel Newton-Krylovapproach (Jones &Woodward, 2001; Kollet & Maxwell, 2006; Maxwell, 2013). ParFlow solves the variably satu-rated flow (i.e., saturated and unsaturated groundwater and surface water) in a single matrix based on hydro-dynamic principles, without a priori specification of the types of flow in certain portion of the domain. Whilethis yields a challenging computational problem, ParFlow takes advantage of a multigrid preconditioner(Ashby & Falgout, 1996) and parallel computation (Kollet et al., 2010).

The source/sink term qr is expressed as

qr ¼ I � E � T ; (3)

where I is the flux of water infiltrating at the land surface (m/hr), E is soil evaporation (m/hr), and T is transpira-tion (m/hr). E is calculated using the Penmann-Monteith equation, with the conductance term assumed to beproportional to soil water content using an analog for Darcy’s law (Millar et al., 2017). T is solved by the par-simonious plant hydraulics scheme in TREES, which predicts a steady-state relation between T and xylempressure (Pc) at a given soil water potential (Ps; equation (4); Sperry et al., 2016). The decline of hydraulic con-ductance (k) from its maximum (kmax) to more negative P is described using a two-parameter Weibull func-tion, known as vulnerability curves (equation (5)), for xylem components (leaf, stem, and root) and a vanGenuchten function (Van Genuchten, 1980) for the rhizosphere, which refers to the soil around each rootthrough which water moves down a pressure gradient from the bulk soil.

T ¼ ∫PdownPup k Pð ÞdP; (4)

k Pð Þ ¼ kmaxe� P=bð Þc½ �: (5)

Following the hydraulic model of Sperry and Love (Sperry & Love, 2015; Sperry et al., 2016), stomata areassumed to regulate the pressure drop ΔP = Ps� Pc based on the fractional drop in whole plant-soil hydraulicconductance from its maximum (k/kmax):

ΔP ¼ ΔP0k=kmax; (6)

where ΔP0 is the unregulated pressure drop, given by the multiplication of vapor pressure deficit, D, and themaximum diffusive conductance parameter, Gmax. This regulated ΔP yields the regulated values for T. Gmax isalternatively coupled to the photosynthesis routine of TREES to capture the response of diffusive conduc-tance to other factors such as light and temperature, given by

G0max ¼ a

Aca � ci

; (7)

where A is photosynthetic carbon assimilated per unit leaf area based on a photosynthesis model (Farquharet al., 1980); a = 1/1.6 is a molar conversion constant between water vapor and CO2 conductance; ca and ci aremolar fractions of CO2 in the leaf surface and in the leaf intercellular air spaces, respectively. The minimum ofGmax and G0

max is used at every time step. This plant hydraulics routine was implemented as a module inTREES, with the meteorological forcing and other parameters similar to the original TREES model (Mackayet al., 2015). Major model parameters and values used in this study are listed in Table 1.

2.2. Data and Model Setup

We examined the water status of a natural riparian cottonwood forest (Populus angustifolia, Populus deltoides,and native hybrids) along a 3-km river corridor within the Oldman River valley near Lethbridge, Alberta,

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4904

Canada (49.702°N, 112.863°W). This is a semiarid prairie region with streamflows originating in the RockyMountain (Rood et al., 2013). Observations of meteorological variables such as photosynthetically activeradiation, precipitation, temperature, wind speed, vapor pressure deficit, streamflow, and leaf area index,ET, and total water in the top 2.50-m depth of soil were made during both wet (2014) and dry (2015) yearswith contrasting May–September cumulative precipitation (2014: 362 mm; 2015: 181 mm) and averagestreamflow rates (2014: 265 m3/s; 2015: 61 m3/s; Flanagan et al., 2017).

ParFlow-TREES was initiated with parameter values listed in Table 1 and the model domain shown in Figure 3.Branch xylem vulnerability to cavitation was measured and was assumed to be the same for leaf, stem, androot segments. Kmax and Gmax were estimated from observed ET. Rooting depth near the study site wasreported to be around 1.0 m (Rood et al., 2011). ParFlow-TREES has the capacity of prescribing spatially het-erogeneous parameters and those parameters were assigned to the extent that information was available.Land cover types were assigned based on a digital height model generated from airborne lidar data(Hopkinson et al., 2005) that differentiated among trees, soil, and stream. Tree-covered grid cells were simu-lated with plant hydraulics and soil evaporation, soil cells were simulated with only soil evaporation, andstream cells were simulated without ET. The average of ET and soil moisture from cells roughly falling withinthe footprint of the eddy covariance flux tower was used to compare against observations (Flanagan et al.,2017). To simplify computation, a smoothed topography was imposed to distinguish cells representing thestream channel, riverbanks, and floodplains (Figure 2b). All cells had a slope of 0.01 m/km in the directionof the river flow. We focused on modeling the stream-floodplain interaction and assumed a negligible con-tribution from upslope drainage, as this ecosystem was mostly influenced by alluvial groundwater (Roodet al., 2013; Willms et al., 1998). A temporally varying streamflow rate was prescribed at the upstream end(Figure 3a) based on the closest gauge station (water ID station 05AD007, data archived by the Water

Table 1Major Inputs of the Model and Values Used in This Study

Model parameter Baseline values (test values)a Sources

(a) TREES parametersWeibull vulnerability curve parameter [b, c] [2.03, 5.25] ([0.95, 1.02]b) Measured for xylem and assumed to be

the same for roots and leaves; test valueswere extracted from Tyree et al. (1994)

Maximum whole plant hydraulicconductance per leaf area (Kmax)

6.5 mmol·s�1·m�2·MPa (22 mmol·s�1·m�2·MPa)b Estimated

Maximum diffusive conductance towater vapor (Gmax)

0.2 mol·s�1·m�2 (0.4 mol·s�1·m�2·MPa)b Estimated

Rooting depth 1 m (2 m)c Rood et al. (2011); test value was assumed(b) ParFlow parametersDomain 50 × 18 × 20Grid size 60 × 60 × 0.5 mVan Genuchten parameters [a, n] [7.5, 1.89] for top 2 m and [14.5, 2.68]

for bottom 8 mLeij et al. (1996)

Saturated permeability 0.0442 m/hr for top 2 m and 9 m/hrfor bottom 8 m

Fleckenstein et al. (2006)

Saturated soil moisture 0.53 ObservedResidual soil moisture 0.1 ObservedManning’s coefficient 2 × 10�6 hr/m1/3 Assumed(c) Initial and boundary conditionBoundary condition No flow except for the surface allowing

overland flow when the water table risesto the land surface

Initial condition Spin-up for one growing season usingthe meteorological forcing and streamflowobserved for the wet year 2014 without simulating trees

Note. ParFlow = Parallel Flow; TREES = Terrestrial Regional Ecosystem Exchange Simulator.aBaseline values were used to compare against observations at the study site; test values were used to evaluate its influence over model output. bFor the xylemvulnerability test; the higher xylem vulnerability was prescribed along with higher Kmax and Gmax values.

cFor the deeper rooting depth test.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4905

Survey of Environment Canada), situated at the downstream edge of the study domain. Measured growingseason meteorological forcing from 1 May to 3 September for a total of 125 days in both years was applieduniformly across the domain.

2.3. Numerical Experiments

ParFlow-TREES provided a framework where mechanisms were isolated and evaluated experimentally bymanipulating model inputs. Here we focused on two hypotheses (Table 2). First, we assessed if increasingplant water sources could buffer drought stress for the ecosystem by adjusting soil permeability parameter-ization and incoming streamflow (Table 2 and Figures 3c–3f). RootZone, Vert_GW, 3D_GW, and3D_GW_Stream represented progressively increasing water sources that plants can access. Using these con-figurations, we imposed an extreme drought that eliminated precipitation from the 2015meteorological con-ditions and repeated for 10 growing seasons (May–September). The observed 2015 streamflow (average rate:55 m3/s) was multiplied by different factors to create a wide range of flow regimes (Table 2). We focused ongrowing seasons based on the strong association between groundwater and cottonwood growth during thegrowing season (Rood et al., 2013; Shepherd et al., 2010). Although this ignored the potential replenishmentof groundwater storage during winter and early spring, it represented an extreme scenario with carry-overeffects of sustained drought between years. To characterize the risk of drought-induced mortality, wereported the average percentage loss of whole-plant hydraulic conductance (PLK) during the growing seasonfor every year. We also reported the percentage of modeled tree cells across the landscape experiencing aPLK greater than 60% for each year, based on previous studies suggesting that 60% is a starting point,beyond which the probability of mortality increased dramatically (Adams et al., 2017; McDowell et al., 2013).

Figure 3. Model setup in the planar (a) and profile (b) directions. The model domain was discretized into 50 × 18 × 20 cellswith a uniform cell dimension of 60 × 60 × 0.5 m, covering a river corridor area of 3 by 1.08 km. In (a), colors represented theland cover types of the model, including trees, soils, and stream channels. Dark green cells represented trees fallingwithin the footprint of the eddy covariance flux tower (red triangle). Streamflow rates were prescribed at the upstream enddesignated by the dark blue arrow. In (b), the vertical domain was discretized to 20 layers for a total depth of 10 m. The top2 m represented a finer textured substrate of sandy loam, and the bottom 8 m represented a coarse substrate of sandand gravel. Riverbanks have a profile slope of 0.03 m/m, and floodplains have a profile slope of 0.004 m/m. All cells had aslope of 0.01 m/km in the river flow direction. Four characterizations of plant water sources (c–f) were compared usingmodel simulations. (c) The 3D_GW_Stream simulation where three-dimensional water flow was solved. (d) The 3D_GWsimulation where streamflow was eliminated from the simulation, while soil parameters were the same as 3D_GW. (e) TheVert_GW simulation where groundwater flows only in the vertical direction by setting lateral soil permeability to zero.(f) The RootZone simulation where trees rely solely on root zone storage by further setting soil permeability beyond theroot zone to zero. Blue arrows within (c)–(f) represent the water flows.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4906

To evaluate the second hypothesis, we repeated the first experiment,but with different sets of plant hydraulic traits including more vulner-able xylem and deeper rooting depth (Table 2). We expect that a morevulnerable xylem would make the system more sensitive to droughtand changes in water sources and a deeper rooting depth would makethe system less sensitive. The impact of other factors, such as topogra-phy, substrate texture and permeability, and vegetation distribution,has been previously reported (Atchley & Maxwell, 2011; Rihani et al.,2010). Although not exhaustive, these numerical experiments providesome insights about the system and address the following questions:(1) How do different groundwater processes buffer plant water stressduring drought? (2) What magnitudes of streamflow sustain current cot-tonwood forests during drought? (3) How do different water sourcesand plant hydraulic traits interact to influence forest sensitivity todrought-induced mortality?

3. Results3.1. Model Evaluation in the Wet and Dry Years

Parflow-TREES with the 3D_GW_Stream scenario reasonably captured the dynamics of ET and soil water bothin the wet (2014) and dry (2015) years (Figure 4). In 2014, soil water was recharged by high streamflow in theearly growing season, whereas in 2015 soil water declined monotonically. In spite of the much lower preci-pitation in year 2015, the total ET was roughly comparable in both years.

3.2. Sensitivity to Multiyear Extreme Drought

As drought progressed over time, PLK increased and the seasonal total ET decreased (Figure 5). But differentsensitivities were predicted when considering alternative plant water sources. When solely relying on waterstored in the root zone (Figure 5, RootZone, red lines), trees experienced a mean PLK of 50% in the first yearand over 90% in subsequent years. Groundwater subsidy provided a buffer against drought stress, but

Table 2Description of Hypotheses and Corresponding Setup of the Numerical Experiments

HypothesisGroundwater

setupPlant

parameters

Streamflowmultiplying

factor

Increasing plant watersources buffers drought stress

3D_GW_Stream Baseline 0.1, 0.5, 1, 23D_GW Baseline 0Vert_GW Baseline 0RootZone Baseline 0

Plant hydraulic traits mediatesensitivity to water sourcesand water supply

RootZone,Vert_GW,3D_GW,3D_GW_Stream

Vulnerablexylem

0.0, 0.1, 0.5,1, 2

Deep roots

Note. 3D_GW_Stream denoted the model setup solving the full three-dimen-sional water flow; 3D_GW represented lateral groundwater flow driven bytopography only; Vert_GW denoted the model setup that only allowed verti-cal flow; RootZone denoted the model setup that solved the flow in the ver-tical direction and within the root zone.

Figure 4. Simulated (solid lines) and observed (solid dots) daily soil water (a, b) and ET (c, d) versus date of year in 2014 (a, c)and 2015 (b, d). Soil water is presented as the total water within the top 2.5 m of soil. R2 of observations versus the mean ofpredictions from model cells falling within the footprint of the eddy covariance measurements is presented.ET = evapotranspiration.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4907

differences could be observed between the predictions of Vert_GWand 3D_GW (Figure 5, green versus black lines). The Vert_GW setupthat only considered the vertical flow predicted lower PLK in the first3 years and higher PLK in subsequent years.

Assuming that tree mortality occurred when the associated PLKexceeded 60%, the RootZone setup predicted that all trees would likelyhave died by the second year of drought. The Vert_GW setup predictednomortality within the first 3 years of drought and 100%mortality afterthe fourth year (Figure 5a, red and green dots). By contrast, the 3D_GWpredicted mortality to progress gradually over time (Figure 5a, blackdots), starting with trees located furthest from the streamchannel (Figure 6).

Without streamflow, groundwater flow was mostly governed by topo-graphic gradients from high to low elevation. Trees located at highertopographic positions lost water to trees at lower positions, experi-enced a deeper water table, and had higher PLK even during the first3 years of drought (Figure 7). As drought sustained, more trees becamestressed with increasing PLK. But a small fraction of trees at lower ele-vation near the stream channel remained buffered from drought stress.Without lateral water redistribution, the Vert_GW model setup pre-dicted little spatial variations in PLK and all trees were dead by the fifthyear of drought (Figure 5).

3.3. Drought Stress Mediated by Streamflow

In the absence of precipitation, higher streamflow raised the watertable and reduced the mean values of PLK (Figure 8), as groundwaterpredominately flowed from stream into the floodplain following thehydraulic gradients. But some trees located away from the stream stillexperienced a PLK greater than 60% (Figures 8 and 9). The simulationssuggested that 6% cottonwoods could be threatened if the streamflowwas to be maintained at the dry year level (multiplier = 1). If the stream-flow was doubled (multiplier = 2), then all trees were buffered fromwater stress even if there was no precipitation.

Figure 5. Predictions of seasonal average PLK (a) and total ET (b) during the10 years of drying down without precipitation and incoming streamflow.Colors represented the three alternate representations of plant water sources bymanipulating the model parameter, saturated permeability. RootZone repre-sented plants taking water solely from root zone storage; Vert_GW representedplants taking water from both root zone and beyond root zone through verticalflow, but lateral flow was ignored; 3D_GW represented plants taking waterfrom the integrated 3-D water flow. The mean values across the landscape werepresented. Solid dots represented the fraction of trees experiencing a PLK over60% across the landscape for a given year. Different dot sizes were used foreasier visualization. ET = evapotranspiration; PLK = percentage loss of hydraulicconductance.

Figure 6. Spatial distribution of seasonal average PLK predicted by the 3D_GW setup, as drought progressed into the (a)first, (b) second, (c) fourth, and (d) tenth years of the extreme drought experiment with no precipitation and no stream-flow. Color gradients represent the magnitude of PLK. Blue cells represented the stream channel. The fractions of tree cellsexperiencing a PLK over 60% for a given year were presented as numbers in the top of each snapshot. PLK = percentageloss of hydraulic conductance.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4908

3.4. Integrated Effects of Relevant Processes in PredictingSensitivity to Drought

The two-way lateral redistribution of groundwater from topographichigh to low and from the stream into the floodplain zone was shownto buffer cottonwood against sustained severe drought (Figure 10,orange lines). This mediating influence was negligible if the droughtwas limited to a single growing season when the alluvial aquifer sto-rage was still high (Figure 10, light blue line). As expected, plant hydrau-lic traits mediated the sensitivity to water sources (Figure 10, differentsymbols). Given the same water supply, cottonwoods with vulnerablexylem were at greater risk of mortality, and growing deeper rootsreduced the risk of mortality.

4. Discussion4.1. Model Evaluation in the Wet and Dry Years

The much lower precipitation in the dry year 2015 had little impact onET (Figure 4), indicating that plants were taking water from sourcesother than precipitation. This has been attributed to the large storagecapacity of the floodplain substrate (Flanagan et al., 2017) and the rela-

tively shallow water table sustained by streamflow (Rood et al., 2013; Scott et al., 1999). Storage could effec-tively buffer cottonwoods against isolated dry years, although reduced growth has been reported when thedrought was severe (Schook et al., 2016).

4.2. Sensitivity to Multiyear Extreme Drought

In spite of the extremely dry conditions and the highly vulnerable xylem of cottonwood (Tyree et al., 1994),3D_GW predicted a gradual increase of mortality risk over time (Figures 5 and 6). The predicted pattern fellwithin the realm of previously documented cottonwood declines at other sites suggesting that mature cot-tonwoods were resilient (Andersen, 2016; Braatne et al., 2007), and that cottonwood decline was gradual,spanning several years to decades (Rood et al., 2003, 1995). Trees located lower in elevation received subsidyat the expense of sacrificing trees at higher positions (Figure 7), similar to the study at a headwater catchment(Shen et al., 2013). The fraction of trees receiving subsidy became smaller over time (Figure 6), resulting in apattern of narrowing forest bands on the floodplain that has been reported (Cordes et al., 1997; Rood et al.,2003). High PLK was predicted when the water table fell beyond a depth of 3 m (Figure 7). This agreed withthe previous findings that cottonwoods were generally restricted to streamside bands where depth to thewater table was no greater than 3.5 m (Busch et al., 1992; Horton et al., 2001; Scott et al., 1999), although cot-tonwoods can survive at places where the water table is much deeper (Zimmermann, 1969).

Figure 7. Temporal means of WTD versus seasonal PLK for individual trees simu-lated in the extreme drought experiment during the first 3 years of drought(open circles) and during the latter 7 years of drought (crosses). Every dotrepresented a single tree cell, and only cells falling within the footprint of theeddy covariance measurements are shown for easier visualization.PLK = percentage loss of hydraulic conductance; WTD = water table depth.

Figure 8. WTD (a) and seasonal average PLK (b) during the 10th growing season with no precipitation and varying magni-tudes of streamflow. The observed streamflow during the dry year (2015) was adjusted by multiplying factors of 0, 0.1, 0.5,1, and 2, respectively. These streamflow scenarios corresponded to a mean discharge rate of 0.0, 6.1, 30.5, 61, and 122m3/s,respectively. The box and whisker plots showed the median, upper, and lower quartiles of the data, with the whiskersextending to the most extreme data point. PLK = percentage loss of hydraulic conductance; WTD = water table depth.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4909

RootZone and Vert_GW predicted a step change from zero to total mortality across the landscape (Figure 5),which is probably unrealistic (Fisher et al., 2017). In particular, the Vert_GW configuration mimics the effectsof a routine that has been widely adopted in current land surface models (M. P. Clark et al., 2015; Fan, 2015).But our results demonstrated that it might underpredict the mortality risk of trees at higher positions early inthe drought and overpredict the mortality risk of trees at lower positions late in the drought. Consequently,this representation of plant water supply could be insufficient for predicting tree survival/mortality atlandscape scales.

4.3. Drought Stress Mediated by Streamflow

Incorporating streamflow (Figures 8 and 9) demonstrated the hydrologic linkage between riparian cotton-woods and stream discharge in semiarid regions where local precipitation is insufficient to support trees(Rood et al., 2013). The spatial average PLK dropped from 73% to less than 20% when streamflow increased

Figure 9. Spatial distribution of seasonal average PLK during the 10th growing season with different magnitudes ofstreamflow that were adjusted by a multiplying factor of (a) 0.1, (b) 0.5, (c) 1, and (d) 2 based on the observed stream-flow during 2015. These scenarios corresponded to a mean discharge rate of 6.1, 30.5, 61, and 122 m3/s, respectively. Colorgradients represent themagnitude of PLK. Blue cells represented the stream channel. The fraction of tree cells experiencinga PLK over 60% for a given year was also presented as numbers in the top of every snapshot. PLK = percentage loss ofhydraulic conductance.

Figure 10. Percentage of tree grid cells experiencing PLK over 60% predicted by different representations of plant watersources from RootZone, Vert_GW, 3D_GW, and stream-floodplain interaction associated with various magnitudes ofstreamflow (Streamflow, numbers represent the multipliers used to rescale the observed streamflow in 2015), for the tenthyear (orange lines) and first year (light blue line) since drought started. Dot symbols represented different plant hydraulictraits being tested. Filled circles represented the baseline model parameter, filled triangles represented cottonwoodswith more vulnerable xylem, and open circles represented cottonwoods with deeper rooting depth. PLK = percentage lossof hydraulic conductance; RootZone = root zone storage; Vert_GW = vertical water flow; 3D_GW = lateral redistributioninduced by topography.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4910

from 0% to 10% of the magnitude in year 2015 (Figure 9b). This is likely to be related to the high transmissiv-ity of riparian substrates characterized by gravels so that water can quickly move into the floodplain (Amlin &Rood, 2003). Notably, higher streamflow was required to replenish places that were distant from the stream(Friedman & Lee, 2002; Leblanc et al., 2012) and to sustain trees growing in these positions with elevation~1 m higher than the riverside trees (Figures 8b and 9d).

Future drought in the headwater catchments could result in thin snowpack and have a profound influence onthe health of downstream riparian forests (Braatne et al., 2007; Rood et al., 1995), especially if it coincides witha reduction in local precipitation. The impact of drought would be enhanced by increased demands for riverwater diversion (Dijk et al., 2013). In turn, riparian ecosystems could be more vulnerable if there was not suffi-cient streamflow to compensate for reduced precipitation and to recharge the local alluvial groundwater.

4.4. Integrated Effects of Relevant Processes in Predicting Sensitivity to Drought

The dependency of tree health on groundwater sources diminished when the drought was short and the sto-rage was high (Figure 10, light blue line). Similarly, the reliance of cottonwoods on streamflow could also bedampened when there is sufficient water supply from local precipitation (Busch et al., 1992; Scott et al., 1999;Snyder & Williams, 2000). Our results have implications for different conceptualizations of plant watersources: They can give similar predictions under wet conditions but dramatically different answers duringsustained extreme drought. Given the anticipated future with more extreme, longer-term or frequentdrought and prolonged low streamflow conditions (Mishra et al., 2010; Sheffield & Wood, 2008), integratingall relevant processes therefore has the potential of being more robust compared to the simpler representa-tions in current models for quantifying the risk of drought-induced forest mortality.

Consistent with studies showing different tolerances to low water potentials across cottonwood genotypes(Kranjcec et al., 1998; Tyree et al., 1994), ParFlow-TREES predicted higher PLK for trees with a more vulnerablexylem under the same hydro-climatic conditions (Figure 10). Trees with deeper roots had greater access tosoil water storage (Johnson et al., 2018) and lower reliance on high streamflow (Figure 10). These results rein-force the growing evidence from both field and modeling studies that accurate evaluation of forest vulner-ability to future drought hinges on the integration of plant hydraulic properties with hydrologicalprocesses (Jackson et al., 2000; Tai et al., 2017).

4.5. Methodological Limitations

Our numerical experiments evaluating themortality risk of a riparian cottonwood forest serve as a useful illus-tration of groundwater controls over plant responses to drought. As with any modeling effort, results comewith a number of limitations.

First, we focused on the magnitude of streamflow and mature cottonwood trees. Extreme flooding eventsand variations in streamflow seasonality have been recognized as important processes for cottonwoodrecruitment and seedling establishment (Braatne et al., 2007; Harner & Stanford, 2003; Lytle & Merritt,2004; Rood et al., 2008; Scott et al., 1997). Furthermore, we used a threshold of 60% seasonal average PLKto report the risk of cottonwood mortality based on the previous studies (Adams et al., 2017; McDowellet al., 2013). But a different value might be needed depending on the various symptoms of decline (e.g.,branch sacrifice, reduced sexual and vegetative reproduction, and slower growth; Rood et al., 2000; Scottet al., 1999), as well as plant traits such as the ability to recover from embolism (Adams et al., 2017;McDowell et al., 2013). Those aspects, although beyond the scope of the current study, should be incorpo-rated to improve predicting the flow requirement for cottonwood restoration.

Second, the model domain was confined to the river and riparian floodplain. Although the study site was pri-marily influenced by alluvial groundwater (Rood et al., 2013; Willms et al., 1998), upslope drainage might con-tribute significant amount of water for plants developing along some rivers (Winter, 1998). Similarly, differentcross-section geometries of the model domain might cause a different distribution pattern of water supply.

Third, we assumed no flux boundary conditions that only allowed water to leave the domain as overland flowthrough the stream channel (Schaller & Fan, 2009) and eliminated potential water exchanges with the dee-per, regional-scale groundwater. Although the same boundary condition was applied to all model simula-tions and should not affect the relative importance of alternate water sources in mediating drought

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4911

impact, it resulted in the model prediction of cottonwood survival after 10 years of no precipitation and nostreamflow (Figure 6d), which was less likely to happen in the real world.

Lastly, although ParFlow-TREES has the capacity to use spatially variable input of hydrologic and plant phy-siological parameters, they were assumed to be uniform in this study due to the lack of spatial information.But heterogeneity in hydraulic conductivity and geomorphic structures in particular has been shown to beimportant for characterizing the river-aquifer (i.e., hyporheic) exchange (Cardenas, 2015; Frei et al., 2009).Furthermore, plants are likely to adapt to local habitat by altering their physiological or morphological traits(L. D. Anderegg & HilleRisLambers, 2015; Bréda et al., 2006; Hacke et al., 2000) over time and/or across space.Improved characterization of the variability in subsurface properties, traits of different individual trees, anddifferent tissues within the same tree will be important to accurately quantify ecosystem sensitivity to futurehydroclimatic conditions. Ongoing efforts at Critical Zone Observatories and global plant trait databasesmight appear to offer promising means for getting such information for subsurface (Billings, 2015; Fan,2015) and plant trait plasticity (Fan et al., 2017; Kattge et al., 2011).

5. Conclusions

We presented ParFlow-TREES, an integrated model that combined recent developments in plant hydraulicsand groundwater hydrology, to mechanistically quantify the heterogeneous forest response to drought atthe landscape scale. Model experiments with a cottonwood riparian forest demonstrated the interplaybetween groundwater processes and plant hydraulics in mediating the risk of mortality under sustaineddrought conditions. ParFlow-TREES also illustrated a mechanistic link between regional-scale streamflowand the ecological consequences of tree survival at the landscape scale. Quantifying their interaction willbe informative for management practices targeting the resilience and/or restoration of riparian ecosystems.

ReferencesAdams, H. D., Zeppel, M. J., Anderegg, W. R., Hartmann, H., Landhäusser, S. M., Tissue, D. T., et al. (2017). A multi-species synthesis of phy-

siological mechanisms in drought-induced tree mortality. Nature Ecology & Evolution, 1(9), 1285–1291. https://doi.org/10.1038/s41559-017-0248-x

Allen, C. D., Breshears, D. D., & McDowell, N. G. (2015). On underestimation of global vulnerability to tree mortality and forest die-off fromhotter drought in the Anthropocene. Ecosphere, 6(8), 1–55.

Amlin, N. M., & Rood, S. B. (2003). Drought stress and recovery of riparian cottonwoods due to water table alteration along Willow Creek,Alberta. Trees, 17(4), 351–358.

Anderegg, L. D., & HilleRisLambers, J. (2015). Drought stress limits the geographic ranges of two tree species via different physiologicalmechanisms. Global Change Biology.

Anderegg, W. R., Flint, A., Huang, C.-y., Flint, L., Berry, J. A., Davis, F. W., et al. (2015). Tree mortality predicted from drought-induced vasculardamage. Nature Geoscience, 8(5), 367–371. https://doi.org/10.1038/ngeo2400

Andersen, D. C. (2016). Flow regime effects on mature Populus fremontii (Fremont cottonwood) productivity on two contrasting drylandriver floodplains. The Southwestern Naturalist, 61(1), 8–17. https://doi.org/10.1894/0038-4909-61.1.8

Ashby, S. F., & Falgout, R. D. (1996). A parallel multigrid preconditioned conjugate gradient algorithm for groundwater flow simulations.Nuclear Science and Engineering, 124(1), 145–159. https://doi.org/10.13182/NSE96-A24230

Atchley, A. L., & Maxwell, R. M. (2011). Influences of subsurface heterogeneity and vegetation cover on soil moisture, surface temperatureand evapotranspiration at hillslope scales. Hydrogeology Journal, 19(2), 289–305. https://doi.org/10.1007/s10040-010-0690-1

Barnett, T. P., Adam, J. C., & Lettenmaier, D. P. (2005). Potential impacts of a warming climate on water availability in snow-dominatedregions. Nature, 438(7066), 303–309. https://doi.org/10.1038/nature04141

Bergstrom, A., Jencso, K., & McGlynn, B. (2016). Spatiotemporal processes that contribute to hydrologic exchange between hillslopes, valleybottoms, and streams. Water Resources Research, 52, 4628–4645. https://doi.org/10.1002/2015WR017972

Beven, K., & Kirkby, M. (1979). A physically based, variable contributing area model of basin hydrology/un modèle à base physique de zoned’appel variable de l’hydrologie du bassin versant. Hydrological Sciences Journal, 24(1), 43–69. https://doi.org/10.1080/02626667909491834

Billings, S. A. (2015). ‘One physical system’: Tansley’s ecosystem as Earth’s critical zone. New Phytologist, 206(3), 900–912.Bond, N. R., Lake, P., & Arthington, A. H. (2008). The impacts of drought on freshwater ecosystems: An Australian perspective. Hydrobiologia,

600(1), 3–16. https://doi.org/10.1007/s10750-008-9326-zBraatne, J. H., Jamieson, R., Gill, K. M., & Rood, S. B. (2007). Instream flows and the decline of riparian cottonwoods along the Yakima River,

Washington, USA. River Research and Applications, 23(3), 247–267. https://doi.org/10.1002/rra.978Bréda, N., Huc, R., Granier, A., & Dreyer, E. (2006). Temperate forest trees and stands under severe drought: A review of ecophysiological

responses, adaptation processes and long-term consequences. Annals of Forest Science, 63(6), 625–644. https://doi.org/10.1051/forest:2006042

Busch, D. E., Ingraham, N. L., & Smith, S. D. (1992). Water uptake in woody riparian phreatophytes of the southwestern United States: A stableisotope study. Ecological Applications, 2(4), 450–459. https://doi.org/10.2307/1941880

Cardenas, M. B. (2015). Hyporheic zone hydrologic science: A historical account of its emergence and a prospectus.Water Resources Research,51, 3601–3616. https://doi.org/10.1002/2015WR017028

Chow, V. T., Maidment, D. R., & Mays, L. W. (1988). Applied hydrology. In R. Eliassen, P. H. King, & R. K. Linsley (Eds.),McGraw-Hill series in waterresources and environmental engineering (572 pp.). McGraw-Hill, Inc.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4912

AcknowledgmentsWe thank Ying Fan from RutgersUniversity and two anonymousreviewers for their comments thatgreatly helped improve the manuscript.We also thank Reed Maxwell fromColorado School of Mines forinformative discussions. This work wasfunded by the National ScienceFoundation (NSF) through grants IOS1450679 and IOS 1450650. This researchwas also part of the Functional Flows: APractical Strategy for Healthy Riversproject and was supported by grantsfrom Alberta Innovates-Energy andEnvironment Solutions, the NaturalSciences and Engineering Council ofCanada - Discovery Grant Program, andConoco Phillips Canada. David Ellis (Cityof Lethbridge) and Coreen Putman(Helen Schuler Nature Centre,Lethbridge) provided permission toconduct research in the Nature Centregrounds. The statements made in thismanuscript reflect the views of theauthors and do not necessarily reflectthe views of the funding agencies. Dataused in this study are available as DataSet S1.

Clark, J. S., Iverson, L., Woodall, C. W., Allen, C. D., Bell, D. M., Bragg, D. C., et al. (2016). The impacts of increasing drought on forest dynamics,structure, and biodiversity in the United States. Global Change Biology, 22(7), 2329–2352. https://doi.org/10.1111/gcb.13160

Clark, M. P., Fan, Y., Lawrence, D. M., Adam, J. C., Bolster, D., Gochis, D. J., et al. (2015). Improving the representation of hydrologic processes inEarth System Models. Water Resources Research, 51(8), 5929–5956. https://doi.org/10.1002/2015WR017096

Cordes, L., Hughes, F., & Getty, M. (1997). Factors affecting the regeneration and distribution of riparian woodlands along a northern PrairieRiver: The Red Deer River, Alberta, Canada. Journal of Biogeography, 24(5), 675–695. https://doi.org/10.1111/j.1365-2699.1997.tb00077.x

Dawson, T. E., & Ehleringer, J. R. (1991). Streamside trees that do not use stream water. Nature, 350(6316), 335–337. https://doi.org/10.1038/350335a0

Dijk, A. I., Beck, H. E., Crosbie, R. S., Jeu, R. A., Liu, Y. Y., Podger, G. M., et al. (2013). The Millennium Drought in Southeast Australia (2001–2009):Natural and human causes and implications for water resources, ecosystems, economy, and society. Water Resources Research, 49(2),1040–1057. https://doi.org/10.1002/wrcr.20123

Dingman, L. S. (1994). Physical hydrology. New York: Macmillan Publishing Company.Fan, Y. (2015). Groundwater in the Earth’s critical zone: Relevance to large-scale patterns and processes. Water Resources Research, 51,

3052–3069. https://doi.org/10.1002/2015WR017037Fan, Y., Miguez-Macho, G., Jobbágy, E. G., Jackson, R. B., & Otero-Casal, C. (2017). Hydrologic regulation of plant rooting depth. Proceedings of

the National Academy of Sciences, 201712381.Fang, Y., Leung, L. R., Duan, Z., Wigmosta, M. S., Maxwell, R. M., Chambers, J. Q., & Tomasella, J. (2017). Influence of landscape heterogeneity

on water available to tropical forests in an Amazonian catchment and implications for modeling drought response. Journal of GeophysicalResearch: Atmospheres, 122, 8410–8426. https://doi.org/10.1002/2017JD027066

Farquhar, G., von Caemmerer, S. V., & Berry, J. (1980). A biochemical model of photosynthetic CO2 assimilation in leaves of C3 species. Planta,149(1), 78–90. https://doi.org/10.1007/BF00386231

Fisher, R. A., Koven, C. D., Anderegg, W. R., Christoffersen, B. O., Dietze, M. C., Farrior, C., et al. (2017). Vegetation demographics in Earth SystemModels: A review of progress and priorities. Global Change Biology, 24(1), 35–54.

Flanagan, L. B., Orchard, T. E., Logie, G. S., Coburn, C. A., & Rood, S. B. (2017). Water use in a riparian cottonwood ecosystem: Eddy covariancemeasurements and scaling along a river corridor. Agricultural and Forest Meteorology, 232, 332–348. https://doi.org/10.1016/j.agrformet.2016.08.024

Fleckenstein, J. H., Niswonger, R. G., & Fogg, G. E. (2006). River-aquifer interactions, geologic heterogeneity, and low-flow management.Groundwater, 44(6), 837–852. https://doi.org/10.1111/j.1745-6584.2006.00190.x

Frei, S., Fleckenstein, J., Kollet, S., & Maxwell, R. (2009). Patterns and dynamics of river–aquifer exchange with variably-saturated flow using afully-coupled model. Journal of Hydrology, 375(3-4), 383–393. https://doi.org/10.1016/j.jhydrol.2009.06.038

Friedman, J. M., & Lee, V. J. (2002). Extreme floods, channel change, and riparian forests along ephemeral streams. Ecological Monographs,72(3), 409–425. https://doi.org/10.1890/0012-9615(2002)072[0409:EFCCAR]2.0.CO;2

Hacke, U., Sperry, J., Ewers, B., Ellsworth, D., Schäfer, K., & Oren, R. (2000). Influence of soil porosity on water use in Pinus taeda. Oecologia,124(4), 495–505. https://doi.org/10.1007/PL00008875

Hanson, P. J., Amthor, J. S., Wullschleger, S. D., Wilson, K., Grant, R. F., Hartley, A., et al. (2004). Oak forest carbon and water simulations: Modelintercomparisons and evaluations against independent data. Ecological Monographs, 74(3), 443–489. https://doi.org/10.1890/03-4049

Harner, M. J., & Stanford, J. A. (2003). Differences in cottonwood growth between a losing and a gaining reach of an alluvial floodplain.Ecology, 84(6), 1453–1458. https://doi.org/10.1890/0012-9658(2003)084[1453:DICGBA]2.0.CO;2

Hopkinson, C., Chasmer, L. E., Sass, G., Creed, I. F., Sitar, M., Kalbfleisch, W., & Treitz, P. (2005). Vegetation class dependent errors in lidarground elevation and canopy height estimates in a boreal wetland environment. Canadian Journal of Remote Sensing, 31(2), 191–206.https://doi.org/10.5589/m05-007

Horton, J., Kolb, T., & Hart, S. (2001). Responses of riparian trees to interannual variation in ground water depth in a semi-arid river basin.Plant, Cell & Environment, 24(3), 293–304. https://doi.org/10.1046/j.1365-3040.2001.00681.x

Jackson, R. B., Sperry, J. S., & Dawson, T. E. (2000). Root water uptake and transport: Using physiological processes in global predictions.Trends in Plant Science, 5(11), 482–488. https://doi.org/10.1016/S1360-1385(00)01766-0

Johnson, D. M., Domec, J. C., Berry, Z. C., Schwantes, A. M., Woodruff, D. R., McCulloh, K. A., et al. (2018). Co-occurring woody species havediverse hydraulic strategies and mortality rates during an extreme drought. Plant, Cell & Environment, 41(3), 576–588. https://doi.org/10.1111/pce.13121

Jones, J. E., & Woodward, C. S. (2001). Newton–Krylov-multigrid solvers for large-scale, highly heterogeneous, variably saturated flow pro-blems. Advances in Water Resources, 24(7), 763–774. https://doi.org/10.1016/S0309-1708(00)00075-0

Kattge, J., Diaz, S., Lavorel, S., Prentice, I. C., Leadley, P., Bönisch, G., et al. (2011). TRY—A global database of plant traits. Global Change Biology,17(9), 2905–2935. https://doi.org/10.1111/j.1365-2486.2011.02451.x

Kollet, S. J., & Maxwell, R. M. (2006). Integrated surface–groundwater flow modeling: A free-surface overland flow boundary condition in aparallel groundwater flow model. Advances in Water Resources, 29(7), 945–958. https://doi.org/10.1016/j.advwatres.2005.08.006

Kollet, S. J., & Maxwell, R. M. (2008). Capturing the influence of groundwater dynamics on land surface processes using an integrated, dis-tributed watershed model. Water Resources Research, 44(2), W02402. https://doi.org/10.1029/2007WR006004

Kollet, S. J., Maxwell, R. M., Woodward, C. S., Smith, S., Vanderborght, J., Vereecken, H., & Simmer, C. (2010). Proof of concept of regional scalehydrologic simulations at hydrologic resolution utilizing massively parallel computer resources. Water Resources Research, 46, W04201.https://doi.org/10.1029/2009WR008730

Kranjcec, J., Mahoney, J. M., & Rood, S. B. (1998). The responses of three riparian cottonwood species to water table decline. Forest Ecologyand Management, 110(1-3), 77–87. https://doi.org/10.1016/S0378-1127(98)00276-X

Krause, S., Lewandowski, J., Grimm, N. B., Hannah, D. M., Pinay, G., McDonald, K., et al. (2017). Ecohydrological interfaces as hotspots ofecosystem processes. Water Resources Research, 53(8), 6359–6376. https://doi.org/10.1002/2016WR019516

Leblanc, M., Tweed, S., Van Dijk, A., & Timbal, B. (2012). A review of historic and future hydrological changes in the Murray-Darling Basin.Global and Planetary Change, 80, 226–246.

Leij, F. J., Alves, W. J., and van Genuchten, M. T. (1996). The UNSODA unsaturated soil hydraulic database: User’s manual, National RiskManagement Research Laboratory, Office of Research and Development, US Environmental Protection Agency.

Lytle, D. A., & Merritt, D. M. (2004). Hydrologic regimes and riparian forests: A structured population model for cottonwood. Ecology, 85(9),2493–2503. https://doi.org/10.1890/04-0282

Mackay, D. S., Roberts, D. E., Ewers, B. E., Sperry, J. S., McDowell, N. G., & Pockman, W. T. (2015). Interdependence of chronic hydraulic dys-function and canopy processes can improve integrated models of tree response to drought. Water Resources Research, 51, 6156–6176.https://doi.org/10.1002/2015WR017244

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4913

Manzoni, S., Vico, G., Porporato, A., & Katul, G. (2013). Biological constraints on water transport in the soil–plant–atmosphere system.Advances in Water Resources, 51, 292–304. https://doi.org/10.1016/j.advwatres.2012.03.016

Maxwell, R. M. (2013). A terrain-following grid transform and preconditioner for parallel, large-scale, integrated hydrologic modeling.Advances in Water Resources, 53, 109–117. https://doi.org/10.1016/j.advwatres.2012.10.001

Maxwell, R. M., Condon, L., & Kollet, S. (2015). A high-resolution simulation of groundwater and surface water over most of the continental USwith the integrated hydrologicmodel ParFlow v3.GeoscientificModel Development, 8(3), 923–937. https://doi.org/10.5194/gmd-8-923-2015

Maxwell, R. M., & Kollet, S. J. (2008). Interdependence of groundwater dynamics and land-energy feedbacks under climate change. NatureGeoscience, 1(10), 665–669. https://doi.org/10.1038/ngeo315

McDowell, N. G., Fisher, R. A., Xu, C., Domec, J., Hölttä, T., Mackay, D. S., et al. (2013). Evaluating theories of drought-induced vegetationmortality using a multimodel–experiment framework. New Phytologist, 200(2), 304–321. https://doi.org/10.1111/nph.12465

Miguez-Macho, G., & Fan, Y. (2012). The role of groundwater in the Amazon water cycle: 2. Influence on seasonal soil moisture and evapo-transpiration. Journal of Geophysical Research, 117(D15), D15114. https://doi.org/10.1029/2012JD017540

Millar, D. J., Ewers, B. E., Mackay, D. S., Peckham, S., Reed, D. E., & Sekoni, A. (2017). Improving ecosystem-scale modeling of evapotran-spiration using ecological mechanisms that account for compensatory responses following disturbance. Water Resources Research, 53,7853–7868. https://doi.org/10.1002/2017WR020823

Mishra, V., Cherkauer, K. A., & Shukla, S. (2010). Assessment of drought due to historic climate variability and projected future climate changein the midwestern United States. Journal of Hydrometeorology, 11(1), 46–68. https://doi.org/10.1175/2009JHM1156.1

Newman, B. D., Wilcox, B. P., Archer, S. R., Breshears, D. D., Dahm, C. N., Duffy, C. J., et al. (2006). Ecohydrology of water-limited environments:A scientific vision. Water Resources Research, 42, W06302. https://doi.org/10.1029/2005WR004141

Phillips, R. P., Ibáñez, I., D’Orangeville, L., Hanson, P. J., Ryan, M. G., & McDowell, N. G. (2016). A belowground perspective on the droughtsensitivity of forests: Towards improved understanding and simulation. Forest Ecology and Management, 380, 309–320. https://doi.org/10.1016/j.foreco.2016.08.043

Powell, T. L., Galbraith, D. R., Christoffersen, B. O., Harper, A., Imbuzeiro, H., Rowland, L., et al. (2013). Confronting model predictions of carbonfluxes with measurements of Amazon forests subjected to experimental drought. New Phytologist, 200(2), 350–365. https://doi.org/10.1111/nph.12390

Richards, L. A. (1931). Capillary conduction of liquids through porous mediums. Physics, 1(5), 318–333. https://doi.org/10.1063/1.1745010Rihani, J. F., Maxwell, R. M., & Chow, F. K. (2010). Coupling groundwater and land surface processes: Idealized simulations to identify effects of

terrain and subsurface heterogeneity on land surface energy fluxes. Water Resources Research, 46, W12523. https://doi.org/10.1029/2010WR009111

Rood, S. B., Ball, D. J., Gill, K. M., Kaluthota, S., Letts, M. G., & Pearce, D. W. (2013). Hydrologic linkages between a climate oscillation, river flows,growth, and wood Δ

13C of male and female cottonwood trees. Plant, Cell & Environment, 36(5), 984–993. https://doi.org/10.1111/

pce.12031Rood, S. B., Bigelow, S. G., & Hall, A. A. (2011). Root architecture of riparian trees: River cut-banks provide natural hydraulic excavation,

revealing that cottonwoods are facultative phreatophytes. Trees, 25(5), 907–917. https://doi.org/10.1007/s00468-011-0565-7Rood, S. B., Braatne, J. H., & Hughes, F. M. (2003). Ecophysiology of riparian cottonwoods: Stream flow dependency, water relations and

restoration. Tree Physiology, 23(16), 1113–1124. https://doi.org/10.1093/treephys/23.16.1113Rood, S. B., Mahoney, J. M., Reid, D. E., & Zilm, L. (1995). Instream flows and the decline of riparian cottonwoods along the St. Mary River,

Alberta. Canadian Journal of Botany, 73(8), 1250–1260. https://doi.org/10.1139/b95-136Rood, S. B., Pan, J., Gill, K. M., Franks, C. G., Samuelson, G. M., & Shepherd, A. (2008). Declining summer flows of Rocky Mountain Rivers:

Changing seasonal hydrology and probable impacts on floodplain forests. Journal of Hydrology, 349(3-4), 397–410. https://doi.org/10.1016/j.jhydrol.2007.11.012

Rood, S. B., Patiño, S., Coombs, K., & Tyree, M. T. (2000). Branch sacrifice: Cavitation-associated drought adaptation of riparian cottonwoods.Trees, 14(5), 0248–0257. https://doi.org/10.1007/s004680050010

Sack, L., Ball, M. C., Brodersen, C., Davis, S. D., Des Marais, D. L., Donovan, L. A., et al. (2016). Plant hydraulics as a central hub integrating plantand ecosystem function: Meeting report for ‘Emerging Frontiers in Plant Hydraulics’(Washington, DC, May 2015). Plant, Cell & Environment,39(9), 2085–2094. https://doi.org/10.1111/pce.12732

Schaller, M. F., & Fan, Y. (2009). River basins as groundwater exporters and importers: Implications for water cycle and climate modeling.Journal of Geophysical Research, 114, D04103. https://doi.org/10.1029/2008JD010636

Schook, D. M., Friedman, J. M., & Rathburn, S. L. (2016). Flow reconstructions in the Upper Missouri River Basin using riparian tree rings.WaterResources Research, 52, 8159–8173. https://doi.org/10.1002/2016WR018845

Scott, M. L., Auble, G. T., & Friedman, J. M. (1997). Flood dependency of cottonwood establishment along the Missouri River, Montana, USA.Ecological Applications, 7(2), 677–690. https://doi.org/10.1890/1051-0761(1997)007[0677:FDOCEA]2.0.CO;2

Scott, M. L., Shafroth, P. B., & Auble, G. T. (1999). Responses of riparian cottonwoods to alluvial water table declines. EnvironmentalManagement, 23(3), 347–358. https://doi.org/10.1007/s002679900191

Sheffield, J., & Wood, E. F. (2008). Projected changes in drought occurrence under future global warming from multi-model, multi-scenario,IPCC AR4 simulations. Climate Dynamics, 31(1), 79–105. https://doi.org/10.1007/s00382-007-0340-z

Shen, C., Niu, J., & Phanikumar, M. S. (2013). Evaluating controls on coupled hydrologic and vegetation dynamics in a humid continentalclimate watershed using a subsurface-land surface processes model. Water Resources Research, 49, 2552–2572. https://doi.org/10.1002/wrcr.20189

Shepherd, A., Gill, K. M., & Rood, S. B. (2010). Climate change and future flows of Rocky Mountain Rivers: Converging forecasts from empiricaltrend projection and down-scaled global circulation modelling. Hydrological Processes, 24(26), 3864–3877. https://doi.org/10.1002/hyp.7818

Snyder, K. A., & Williams, D. G. (2000). Water sources used by riparian trees varies among stream types on the San Pedro River, Arizona.Agricultural and Forest Meteorology, 105(1-3), 227–240. https://doi.org/10.1016/S0168-1923(00)00193-3

Sperry, J. S., & Love, D. M. (2015). What plant hydraulics can tell us about responses to climate-change droughts. New Phytologist, 207(1),14–27. https://doi.org/10.1111/nph.13354

Sperry, J. S., Wang, Y., Wolfe, B. T., Mackay, D. S., Anderegg, W. R. L., McDowell, N. G., & Pockman, W. T. (2016). Pragmatic hydraulic theorypredicts stomatal responses to climatic water deficits. The New Phytologist, 212(3), 577–589. https://doi.org/10.1111/nph.14059

Swetnam, T. L., Brooks, P. D., Barnard, H. R., Harpold, A. A., & Gallo, E. L. (2017). Topographically driven differences in energy and waterconstrain climatic control on forest carbon sequestration. Ecosphere, 8(4). https://doi.org/10.1002/ecs2.1797

Tai, X., Mackay, D. S., Anderegg, W. R., Sperry, J. S., & Brooks, P. D. (2017). Plant hydraulics improves and topography mediates prediction ofaspen mortality in southwestern USA. New Phytologist, 213(1), 113–127. https://doi.org/10.1111/nph.14098

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4914

Taylor, R. G., Scanlon, B., Döll, P., Rodell, M., Van Beek, R., Wada, Y., et al. (2013). Ground water and climate change. Nature Climate Change,3(4), 322–329. https://doi.org/10.1038/nclimate1744

Thompson, S. E., Harman, C. J., Troch, P. A., Brooks, P. D., & Sivapalan, M. (2011). Spatial scale dependence of ecohydrologically mediatedwater balance partitioning: A synthesis framework for catchment ecohydrology. Water Resources Research, 47, W00J03. https://doi.org/10.1029/2010WR009998

Toth, J. (1963). A theoretical analysis of groundwater flow in small drainage basins. Journal of Geophysical Research, 68(16), 4795–4812.https://doi.org/10.1029/JZ068i016p04795

Tyree, M. T., Kolb, K. J., Rood, S. B., & Patiño, S. (1994). Vulnerability to drought-induced cavitation of riparian cottonwoods in Alberta: Apossible factor in the decline of the ecosystem? Tree Physiology, 14(5), 455–466. https://doi.org/10.1093/treephys/14.5.455

Van Genuchten, M. T. (1980). A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society ofAmerica Journal, 44(5), 892–898. https://doi.org/10.2136/sssaj1980.03615995004400050002x

Venturas, M. D., Sperry, J. S., & Hacke, U. G. (2017). Plant xylem hydraulics: What we understand, current research, and future challenges.Journal of Integrative Plant Biology, 59(6), 356–389. https://doi.org/10.1111/jipb.12534

Western, A. W., Grayson, R. B., Blöschl, G., Willgoose, G. R., & McMahon, T. A. (1999). Observed spatial organization of soil moisture and itsrelation to terrain indices. Water Resources Research, 35(3), 797–810. https://doi.org/10.1029/1998WR900065

Willms, J., Rood, S. B., Willms, W., & Tyree, M. (1998). Branch growth of riparian cottonwoods: A hydrologically sensitive dendrochronologicaltool. Trees, 12(4), 215–223. https://doi.org/10.1007/s004680050143

Winter, T. C. (1998). Ground water and surface water: A single resource. U.S. Geological Survey Circular 1139. Denver, CO: U.S. governmentprinting office.

Xu, C., McDowell, N. G., Sevanto, S., & Fisher, R. A. (2013). Our limited ability to predict vegetation dynamics under water stress. NewPhytologist, 200(2), 298–300. https://doi.org/10.1111/nph.12450

Zimmermann, R. C. (1969). Plant ecology of an arid basin, Tres Alamos-Redington area, southeastern Arizona. Rep. 2330-7102,U.S Government Printing Office.

10.1029/2018WR022801Water Resources Research

TAI ET AL. 4915